Implement ISO8601 durations as addition for Offset Aliases

[Haigutus]

Code Sample, a copy-pastable example if possible

current     -> query_config["aggregation"] = "5min"
proposed -> query_config["aggregation"] = "PT5M"

resampled_dataframe = dataframe.resample(query_config["aggregation"], how = query_config["aggregation_type"], closed = query_config["closed"])

Problem description

https://en.wikipedia.org/wiki/ISO_8601#Durations

Current behavior uses non standardized markup for time duration, wich can cause errors and confusion. Implementing ISO8601 duration grantees standardized solution for time duration in PANDAS. It does not have to replace the current system rather extend it.

[mroeschke]

I think this is a reasonable suggestion.

IIRC @WillAyd allowed to_timedelta to parse these durations. Some careful consideration will need to be address regarding how they are mapped to our Tick/DateOffset classes and existing offset aliases.

[WillAyd]

Agreed this makes sense. Support for to_timedelta was added in #21933 so can certainly leverage that as a reference for potential implementation. PRs welcome

[Haigutus]

Maybe this could be reused - https://bitbucket.org/nielsenb/aniso8601

[Flix6x]

Compatibility with ISO8601 durations may not be possible before #22864 is resolved, because "P1D" (an ISO8601 nominal duration of 1 day) denotes a calendar day rather than 24 hours.